This collection explores the results of asking a text-based Large Language Model (LLM) to generate ASCII pictures in response to a range of simple prompts. In subverting their expected, normative functioning as text generating systems, the exercise documented in this collection suggests other expressive potentials LLMs might possess. In particular, the excavation of unusual tendencies within their default operating modes highlights the capacity of these systems to yield strangeness and surprise along a distinctly more-than-human vector. Rather than being ‘glitch art’, the highly abstract images generated speak back to the long established interplay between text as a communicative vehicle and as a concrete visual formation, but in terms that can only be discerned (or, indeed, contextualised) from an algorithmic standpoint. Thus, while the images are ‘meaningless’, in any denotative sense, they invite reflection anew on the hold textual forms have on the world they articulate, as animated by ever-shifting technologies of information inscription and retrieval.
The text of Text-Image can be viewed by clicking on the page spread below, or through this link.